Merge pull request #336 from opencompl/sasha/new-ci

Author: superlopuh

.github/workflows/ci-pivoted.yml

@@ -19,20 +19,20 @@ jobs:
   run-experiments:
     runs-on: ubuntu-latest
     container:
-      image: ghcr.io/opencompl/snitch-toolchain:2.10
+      image: ghcr.io/opencompl/snitch-toolchain:4.0.0
       options: --volume ${{ github.workspace }}:/src
     steps:
       - uses: actions/checkout@v3
         with:
           submodules: recursive
       - name: Install dependencies in venv
         run: |
-          python3 -m venv venv
-          . venv/bin/activate
-          pip install -r /src/requirements.txt
+          uv venv -p 3.11
+          uv pip install -r /src/requirements.txt
+          uv pip install /opt/snitch_cluster/
       - name: Clean and make pivoted
         run: |
-          . venv/bin/activate
+          . .venv/bin/activate
           cp results/pivoted.csv results/pivoted_old.csv.txt
           make clean
           make

Snakefile

@@ -350,8 +350,8 @@ rule profile_to_csv:
 
 rule trace_to_profile:
     input:
-        txt="kernels/{kernel}/{shape}/{variant}.logs/trace_hart_00000000.trace.txt",
-        json="kernels/{kernel}/{shape}/{variant}.logs/trace_hart_00000000.trace.json",
+        txt="kernels/{kernel}/{shape}/{variant}.logs/trace_hart_00000.trace.txt",
+        json="kernels/{kernel}/{shape}/{variant}.logs/trace_hart_00000.trace.json",
     output:
         "kernels/{kernel}/{shape}/{variant}.profile.json",
     params:
@@ -369,23 +369,23 @@ rule trace_to_profile:
 
 rule dasm_to_trace:
     input:
-        "{test}.logs/trace_hart_00000000.dasm",
-    output:
-        txt="{test}.logs/trace_hart_00000000.trace.txt",
-        json="{test}.logs/trace_hart_00000000.trace.json",
+        "{test}.logs/trace_hart_00000.dasm",
+    log:
+        txt="{test}.logs/trace_hart_00000.trace.txt",
+        json="{test}.logs/trace_hart_00000.trace.json",
     params:
         spike=config["spike"],
         gentrace=config["gentrace"],
     shell:
-        "{params.spike} < {input} | {params.gentrace} --permissive -d {output.json} > {output.txt}"
+        "{params.spike} < {input} | {params.gentrace} --permissive --dump-hart-perf {log.json} -o {log.txt}"
 
 
 # Rule used to generate traces for debugging purposes, not used for csv generation
 rule dasm_to_trace_debug:
     input:
-        "{test}.logs/logs/trace_hart_00000000.dasm",
+        "{test}.logs/logs/trace_hart_00000.dasm",
     output:
-        txt="{test}.logs/logs/trace_hart_00000000.trace.txt",
+        txt="{test}.logs/logs/trace_hart_00000.trace.txt",
     params:
         spike=config["spike"],
         gentrace=config["gentrace"],
@@ -397,7 +397,7 @@ rule verilator:
     input:
         "{test}.x",
     output:
-        expand("{{test}}.logs/trace_hart_0000000{hart}.dasm", hart=range(9)),
+        expand("{{test}}.logs/trace_hart_0000{hart}.dasm", hart=range(9)),
     params:
         workdir="{test}.logs",
         vltsim=config["vltsim"],
@@ -548,7 +548,7 @@ rule cc_compile_c:
         cc=config["cc"],
         cflags=config["cflags"],
     shell:
-        "{params.cc} -I$(dirname {input}) -I$(dirname {output}) {params.cflags} -S -o {output} {input.c}"
+        "{params.cc} -I$(dirname {input}) -I$(dirname {output}) {params.cflags} -S -x c++ -o {output} {input.c}"
 
 
 rule cc_compile_ll:
@@ -615,7 +615,7 @@ rule cc_compile_shared_main:
         cc=config["cc"],
         cflags=config["cflags"],
     shell:
-        "{params.cc} -I$(dirname {input.h}) {params.cflags} -S -o {output} {input.c}"
+        "{params.cc} -I$(dirname {input.h}) {params.cflags} -S -x c++ -o {output} {input.c}"
 
 
 ###########################################################

default.yaml

@@ -2,7 +2,7 @@
 xdsl-opt: xdsl-opt # from virtualenv
 mlir-opt: /usr/bin/mlir-opt-16
 mlir-translate: /usr/bin/mlir-translate-16
-cc: /opt/snitch-llvm/bin/riscv32-unknown-elf-clang
+cc: /opt/snitch-llvm/bin/clang++
 spike: /opt/snitch-spike/bin/spike-dasm
 vltsim: /opt/snitch-rtl/bin/snitch_cluster.vlt
 gentrace: /opt/snitch_cluster/util/trace/gen_trace.py
@@ -60,13 +60,14 @@ ldflags:
   - -fuse-ld=/opt/snitch-llvm/bin/ld.lld
   - -nostartfiles
   - -nostdlib
-  - -lclang_rt.builtins-riscv32
-  - -lc
-  - -lsnRuntime
-  - -L/opt/snitch-llvm/lib/clang/12.0.1/lib/
-  - -T/opt/snitch_cluster/sw/snRuntime/base.ld
+  - -L/opt/snitch-llvm/lib/clang/15.0.0/lib/
+  - -L/opt/snitch_cluster/target/snitch_cluster/sw/runtime/
   - -L/opt/snitch_cluster/target/snitch_cluster/sw/runtime/rtl
   - -L/opt/snitch_cluster/target/snitch_cluster/sw/runtime/rtl/build
+  - -T/opt/snitch_cluster/sw/snRuntime/base.ld
+  - -lc
+  - -lsnRuntime
+  - -lclang_rt.builtins-riscv32
 xdsl-passes:
   - test-lower-linalg-to-snitch
 mlir-opt-flags-linalg:

docker/venv.sh

@@ -8,10 +8,10 @@ FORCE_VENV=0
 if [[ -d /src/xdsl ]]; then
   pushd /src/xdsl > /dev/null
   if [[ 1 -eq ${FORCE_VENV} ]] || [[ ! -d ${VENV_DIR} ]] || [[ ! -f ${VENV_DIR}/bin/activate ]]; then
-    python3 -m venv ${VENV_DIR}
+    uv venv -p 3.11 ${VENV_DIR}
     source ${VENV_DIR}/bin/activate
-    pip install --upgrade pip
-    pip install -r /src/requirements.txt
+    uv pip install -r /src/requirements.txt
+    uv pip install /opt/snitch_cluster/
   else
     source ${VENV_DIR}/bin/activate
   fi

kernels/conv2d_d1_s1_3x3/baseline.c.template

@@ -2,8 +2,12 @@
 
 #include <stdint.h>
 
-void conv_2d_nchw_fchw_d1_s1_3x3(const double* restrict x, const double* restrict y,
-                                 double* restrict z) {
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void conv_2d_nchw_fchw_d1_s1_3x3(const double* x, const double* y,
+                                 double* z) {
     for (int row = 0; row < NEW_H; ++row) {
         for (int col = 0; col < NEW_W; ++col) {
             for (int i = 0; i < N; i++) {
@@ -27,3 +31,7 @@ void conv_2d_nchw_fchw_d1_s1_3x3(const double* restrict x, const double* restric
         }
     }
 }
+
+#ifdef __cplusplus
+}
+#endif

kernels/conv2d_d1_s1_3x3/data.h.template

@@ -8,9 +8,17 @@
 #define NEW_H {{M}}
 #define NEW_W {{N}}
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
 extern const double X[N * C * H * W];
 extern const double Y[F * C * 3 * 3];
 extern const double Z_IN[N * F * NEW_H * NEW_W];
 extern const double Z_OUT[N * F * NEW_H * NEW_W];
 
+#ifdef __cplusplus
+}
+#endif
+
 #define TEST_COUNT (N * F * NEW_H * NEW_W)

kernels/conv2d_d1_s1_3x3/main.c

@@ -5,7 +5,7 @@
 #include <math.h>
 
 // Kernel provided via external definition
-void conv_2d_nchw_fchw_d1_s1_3x3(double *x, double *y, double *z);
+extern "C" void conv_2d_nchw_fchw_d1_s1_3x3(double *x, double *y, double *z);
 
 int main() {
     // Allocate shared local memory
@@ -18,9 +18,10 @@ int main() {
 
     // Copy data in shared local memory
     if (snrt_is_dm_core()) {
-        snrt_dma_start_1d(local_x, X, N * C * H * W * sizeof(double));
-        snrt_dma_start_1d(local_y, Y, F * C * 3 * 3 * sizeof(double));
-        snrt_dma_start_1d(local_z, Z_IN, N * F * NEW_H * NEW_W * sizeof(double));
+        snrt_dma_start_1d(local_x, (volatile void *)X, N * C * H * W * sizeof(double));
+        snrt_dma_start_1d(local_y, (volatile void *)Y, F * C * 3 * 3 * sizeof(double));
+        snrt_dma_start_1d(local_z, (volatile void *)Z_IN,
+                          N * F * NEW_H * NEW_W * sizeof(double));
         snrt_dma_wait_all();
     }
 

kernels/ddot/128xf64/baseline.c

@@ -2,10 +2,18 @@
 
 #include <stdint.h>
 
-void ddot(const double* restrict x, const double* restrict y, double* restrict g) {
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void ddot(const double* x, const double* y, double* g) {
     double acc = *g;
     for (uint32_t i = 0; i < N; ++i) {
         acc += x[i] * y[i];
     }
     *g = acc;
 }
+
+#ifdef __cplusplus
+}
+#endif

kernels/ddot/128xf64/data.c

@@ -1,3 +1,5 @@
+#include "data.h"
+
 #define N 128
 
 const double X[N] = {

kernels/ddot/128xf64/snrt.c

@@ -4,18 +4,22 @@
 
 #include <stdint.h>
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
 // x[ N ]
 // y[ N ]
 // g[ 1 ]
-void ddot(const double* restrict x, const double* restrict y, double* restrict g) {
+void ddot(const double* x, const double* y, double* g) {
     snrt_ssr_loop_1d(SNRT_SSR_DM_ALL,
                      // Bounds
                      N,
                      // Strides
                      sizeof(double));
 
-    snrt_ssr_read(SNRT_SSR_DM0, SNRT_SSR_1D, x);
-    snrt_ssr_read(SNRT_SSR_DM1, SNRT_SSR_1D, y);
+    snrt_ssr_read(SNRT_SSR_DM0, SNRT_SSR_1D, (volatile void*)x);
+    snrt_ssr_read(SNRT_SSR_DM1, SNRT_SSR_1D, (volatile void*)y);
 
     snrt_ssr_enable();
 
@@ -32,3 +36,7 @@ void ddot(const double* restrict x, const double* restrict y, double* restrict g
 
     snrt_ssr_disable();
 }
+
+#ifdef __cplusplus
+}
+#endif